I almost missed the "Kill-a-watt" type unit" description you used for
your watt meter because I was rather surprised that an actual Kill-A-Watt
meter could produce such wildly inaccurate readings.
Unfortunately, a lot of the cheaper $10 units that were being sold in
WallMart or ToolMart shops a few years ago left a lot to be desired
regarding accuracy on non linear loads below the 100W mark and were
totally misleading with smaller smpsu type wallwarts where a real 6 watt
load could read either zero or some ficticious 15 to 20 watt figure. I
suspect the model of digital watt meter you're using is one such gadget.
Up until about 12 months ago, I was of the opinion that all such digital
watt meters were prone to this problem (as demonstrated by a couple of
£9.99 meters bought from our local ToolMart shop a few years back).
What changed my mind about the Kill-A-Watt meter's accuracy was my
acquiring a UK version of this meter for a couple of quid from a flea
market trader who specialises in flogging Maplin's returns. The particular
model in question being a Chinese made "Plug-In Mains Power & Energy Usage
Monitor" (as per the box description) made especially for Maplin with the
model code number of L61AQ.
<
http://www.maplin.co.uk/plug-in-mains-power-and-energy-monitor-38343>
Curiously, on opening the box to look at the instruction leaflet, it
appears to be a Prodigit Electronics Co. Ltd product with the model number
of "2000MU-UK". The leaflet is quite clearly referring the meter being
sold by Maplin as an L61AQ.
The power _consumption_ of the meter itself is misquoted on the Maplin
web page as being 20W (which I've also seen misquoted as 10W in the specs
for the Kill-A-Watt P4460). The leaflet more accurately quotes this as 20
VAmax. In reality, checking it with the Metrawatt, the actual power draw,
as best as I can make out using a jeweller's loupe, is just over half a
watt.
The extremely low power factor of 0.025 is exactly what you can expect
when a simple capacitor dropper is used to produce the 5 (or possibly 12)
volt supply required of the processor /ADC chip used in these meters. When
the same circuit is used in a 120 volt device as opposed to a 240 volt
one, the VA figure is halved when delivering the same power to the digital
electronics of the meter.
Going by the PDF of the Kill-A-Watt P4460 manual, it appears to be made
by a different manufacturer using the name P3 International Corp. Since
they're physically identical, apart from the country specific socket,
these items have clearly been made in the same factory, I can only suppose
that "The Global Market Economy" has made manufacturing names as
disposable as email addresses. At the very least, you can bet your bottom
dollar that the L61AQ (aka 2000MU-UK) and the Kill-A-Watt P4460 were both
made in the same Chinese factory.
Whilst the accuracy on sub 10 watt loads is reasonable (around the +/-
15% mark), despite measuring to tenths of a watt, it simply fails to
register the 0.7W consumption of the Metrawatt when switched to the 200v
range (it shows, inaccurately, a reading of 1.7W when the Metrawatt is
switched to the 100v range - the other digital meter ( another Maplin
item, product code N67FU, that I'd bought a few months earlier at the same
flea market) displays this as 1.3W which I'm pretty confident is the more
accurate figure as it's almost twice as much as the 0.7W reading due to
the halving of the multiplier resistance in the Metrawatt when dropping
from the 200v range into the 100v range.
I've just realised that I can more accurately assess the sub 2W accuracy
of these digital meters by measuring the resistance of the MetraWatt
across the live and neutral pins (essentially measuring the voltage
multiplier resistor used on each voltage range). Rather conveniently, it
seems to be based on a notional meter sensitivity of 500 ohms per volt
since I get 50.1K, 100.7K and 247K ohms on the 100, 200 and 500 volt
ranges respectively.
Based on the voltage reading of 246.2v that I was getting from the
2000MU-UK (aka a 240v Kill-A-Watt meter) The other digital watt meter was
only over reading by 0.1W in both cases. The 1.7W reading from a 1.2W load
using the 2000MU-UK represents an error in excess of +40%.
For measuring sub 10W loads, the N67FU walks all over the 2000MU-UK in
terms of accuracy. I believe the 2000MU-UK (aka a 240v Kill-A-Watt meter)
loses out because they seem to have squandered a 4 decimal digit accuracy
to the voltage ADC instead of where such accuracy matters most, i.e. the
Current ADC.
The N67FU can only show voltage to 3 digits so it's reasonable to assume
the reason for it being able to show sub 1 watt readings with a fair
accuracy (as opposed to NO accuracy of the 2000MU-UK in this wattage
range) is that they've used the higher bit accuracy on the current sensor
ADC.
Although I wasn't as impressed by the 2000MU-UK (aka a 240v Kill-A-Watt
meter) when it came to sub 5 watt load measuring accuracy, this was only
on account I'd been even more impressed by the previously acquired N67FU
meter's accuracy in this very low power range. After my experience with
the ToolMart meter, both of these models couldn't fail to impress. ;-)
The only minor deficiency in the N67FU meter is the absence of the PF and
VA measurement functions contrary to these being specified in the list of
functions shown on the packaging. It's only a very minor concern since
it's very easy to calculate these figures, if required, from separate
voltage and current readings (which gets you a VA figure which can then be
used with the real power reading to obtain a PF figure). I have to say,
I've never felt the need to make such measurements.
The bottom line, regarding that useless watt meter of yours is that you
should replace it with an actual Kill-A-Watt (or better) meter if you want
to make reasonably accurate measurements of your computer kit's power
consumption.
HTH & HAND